In 2023, we conducted a national survey to support our narrative change work. This document explores results from that survey related to poverty narratives, including their prevalence and differences across time and geography. It is intended as an internal resource for Narrative Change team members interested in better understanding the analyses we have conducted from this national survey. If you would like to quickly learn our main observations, search “Take-away” in this document or check out the deck we provided Wells Fargo (linked below).

For easy reference, below is a list of other resources or outputs relevant to this work:

  • A purpose document about why we ran this survey
  • The survey questions
  • A data descriptives report, which provides further information about the survey participants and data quality. In addition, basic summary statistics for the non-narrative constructs we measured in the survey can be found in this data descriptives document, as the the stats presented in the present document are intentionally only focused on poverty narratives.
  • We conducted a separate more in-depth analysis on the new psychosocial constructs that were added in this round of the national survey.
  • The R code for the present document
  • External reports
    • Deck for WF on general results from this survey (LINK TO BE ADDED)
    • Segmentation analysis report

Narrative Prevalence

Note: All analyses below are based on raw prevalence data. We have not engaged in any weighting or adjustments for demographic representation.

Individual Questions

question narrative Disagree / Strongly Disagree Neither Agree / Strongly Agree
Welfare makes people lazy. welfare 37% 25% 38%
There is a lot of fraud among welfare recipients. welfare 19% 28% 53%
An able-bodied person collecting welfare is ripping off the system. welfare 27% 26% 47%
Poor people think they deserve to be supported. welfare 34% 34% 32%
Many people take advantage of the welfare system. welfare 19% 21% 60%
Everyone has an equal opportunity to succeed. meritocracy 38% 20% 41%
We live in a meritocracy. Anyone can attain the American Dream. meritocracy 38% 25% 37%
Everyone has an equal opportunity to get a good education. meritocracy 44% 19% 37%
There will always be poor people. fatalism 8% 22% 70%
Almost by definition, someone has to be poor. fatalism 34% 39% 27%
Poverty is an inevitable outcome of society. fatalism 28% 32% 40%
Poor people need guidance to make better life choices. paternalism 20% 34% 46%
There would be less poverty if we helped the poor plan their lives better. paternalism 24% 34% 41%
Low income people could do better if someone helped them spend their money wisely. paternalism 29% 38% 33%
We should teach poor people how to manage their finances. paternalism 13% 34% 53%
Racism makes discrimination against poor minorities worse. structural 13% 21% 66%
Poor people experience prejudice and discrimination in hiring and promotion decisions at work. structural 16% 28% 57%
Poor people lack affordable housing options. structural 11% 17% 72%
Poor people lack affordable child care. structural 11% 19% 70%
Poor people lack opportunities for training & continuing education. structural 23% 23% 54%

Narrative Constructs

narrative mean sd se n
welfare 3.260019 0.9242634 0.0164185 3169
meritocracy 2.971284 1.0909388 0.0193794 3169
fatalism 3.264858 0.7650305 0.0135899 3169
paternalism 3.260177 0.7828676 0.0139068 3169
structural 3.667971 0.7722702 0.0137185 3169

Joint score of harmful narratives (i.e., all but structural): mean = 3.1890844, sd = 0.6378865, se = 0.0113314

Across geography

City

Urban-Rural

According to https://greatdata.com/product/urban-vs-rural the Department of Defense established the following designations for a ZIP Code:
Urban: 3,000+ persons per square mile
Suburban: 1,000 ‐ 3,000 persons per square mile
Rural: less than 1,000 persons per square mile
This isn’t perfect because a lightly populated ZIP Code adjoining a major metropolitan area would be mistakenly classified as a “Rural Area”, but we’ll use it anyway.

US Region

Across time

Comparing these results with those observed in the 2021 national survey known as “pilot1c” (N = 461). Note that in the 2023 national survey, the options were: Strongly disagree, Disagree, Neither agree nor disagree, Agree, Strongly agree. On the other hand, in pilot1c, options were: Strongly agree, Somewhat agree, Neither agree nor disagree, Somewhat disagree, Strongly disagree (in that order). These have been reverse coded appropriately in the analyses below, but that can’t change the fact that the options presented were different.

In addition to the data discussed here, we collected data from an academic collaboration in 2022. In forthcoming analyses we will include that here, which will add a third time point that’s in between the two below.

Individual Questions

Narrative Constructs

T-tests to see if the 2023 average values of each narrative significantly differed from 2021:

narrative mean_2021 mean_2023 diff perc_cng t p.value
welfare 3.195662 3.260019 0.0643573 2% 2.2515863 0.0244230
meritocracy 3.139552 2.971284 -0.1682674 -5% -4.3557411 0.0000140
fatalism 3.167028 3.264858 0.0978293 3% 2.7900253 0.0053291
paternalism 3.400217 3.260177 -0.1400402 -4% -5.0926188 0.0000004
structural 3.711497 3.684443 -0.0270537 -1% -0.8132479 0.4161885

Take-away

All of the harmful narratives showed statistically significant differences from the 2021 pilot1c survey. The structural narrative did not. Endorsement of welfare fraud and fatalism increased, while meritocracy and paternalism decreased. The meritocracy and paternalism effect sizes were notably stronger. We would consider a 4-5% decrease to be a moderate-sized effect. No individual question stands out as driving these effects.

An extremely important caveat to the interpretation of this is that we do not know to what extent these differences reflect true changes over time vs. are the by-product of differential survey uptake. For example, it may be that more people from a certain group who tends to report lower meritocracy views (e.g. low-income non-white women) happened to make up a larger proportion of the 2023 survey than the 2021. The appropriate method to account for this would be to weight the results from various demographic groups to estimate what they would be if representation in both surveys were similar. Given the requirements of such an analysis and competing priorities, this is not something R&E plans to undertake at this time.

For more information about the demographics and representativeness of the 2023 national survey, see the descriptives analysis report. For a comparison to the demographics in the pilot1c 2021 survey, see the admin burden report.

Relationship b/w narratives

Correlation matrices

Stats

r values:

welfare meritocracy fatalism paternalism structural
welfare 1.0000000 0.5561695 0.3455670 0.3738031 -0.3515070
meritocracy 0.5561695 1.0000000 0.1903214 0.2905976 -0.4387403
fatalism 0.3455670 0.1903214 1.0000000 0.2525644 0.0489486
paternalism 0.3738031 0.2905976 0.2525644 1.0000000 0.1110818
structural -0.3515070 -0.4387403 0.0489486 0.1110818 1.0000000

p-values (most are so low that they are difficult to represent computationally, like less than 0.00000000000000022):

welfare meritocracy fatalism paternalism structural
welfare NA 0 0.0000000 0 0.0000000
meritocracy 0 NA 0.0000000 0 0.0000000
fatalism 0 0 NA 0 0.0058501
paternalism 0 0 0.0000000 NA 0.0000000
structural 0 0 0.0058501 0 NA

Take-away

All of these constructs technically have some level of correlation with one another, but the most notable is the relationship between welfare exploitation and meritocracy, which are correlated at r = 0.5561695, meaning that one variable explains 30.93% of the variance in the other. It is also notable that the structural narrative is negatively correlated with these two, but not with fatalism and paternalism.

Relationship w/ new psychosocial constructs

In our previous national survey work, we have examined the relationship between narratives and worldview ideologies (e.g. SDO, RWA, etc.). Although establishing these associations is useful for our understanding, these ideologies aren’t very helpful when it comes to design (see below section for more information about what these ideologies are useful for and why we measure them). We need to look to other psychological theories for direction when it comes to behavior change.

There are many potentially relevant psychological theories and frameworks to draw from. We selected these specific variables by doing a post-mortem on our previous design work. We extracted common themes and objectives across cities by looking at the interventions designed, as well as the LNT defined harmful, authentic, and deep narratives. Next we found variables or theories in the literature that matched the objectives laid out in our theory of change. We also only selected variables we consider modifiable (they had to be changeable using behavior science campaigns or interventions) and evidence-based (there had to be existing examples of behavior change approaches using these theories as active ingredients that we could adapt). Our goal is to strengthen the connective tissue between the existing BSci literature, our ideas42 research, and our narrative change designs.

Below is just a selection of the analyses we have conducted on the new psychosocial constructs. See here for the more detailed analysis.

Social factors

One framework we chose is social change theory, which includes the varialbles:

  • social identity: a person’s knowledge that they belong to certain social groups and that these groups have some cognitive, emotional, and/or valued significance/commonalities
  • social connection: an attribute of the self that reflects cognitions of enduring interpersonal closeness with the social world
  • social norms: the shared rules and expectations within society (or groups) regarding the acceptable behaviors, values, and beliefs of its members

We chose to measure only social identity and social connectedness. This framework represents our relationship with the social ecosystem and how that might explain differences in narrative endorsement.

Basic stats

construct mean SE
social_connection 3.344588 0.011830
social_identity 3.362890 0.016744

Correlations

r values:
Colored according to the following rules of thumb:

  • Strong Relationship: r = ±.5
  • Moderate Relationship: r = ±.3
  • Weak Relationship: r = ±.1
welfare meritocracy fatalism paternalism structural social_identity social_connection
welfare 1.0000 0.5562 0.3456 0.3738 -0.3515 0.0165 -0.0064
meritocracy 0.5562 1.0000 0.1903 0.2906 -0.4387 0.1332 0.0726
fatalism 0.3456 0.1903 1.0000 0.2526 0.0489 0.0234 -0.0971
paternalism 0.3738 0.2906 0.2526 1.0000 0.1111 0.1381 0.0461
structural -0.3515 -0.4387 0.0489 0.1111 1.0000 0.0429 -0.0036
social_identity 0.0165 0.1332 0.0234 0.1381 0.0429 1 0.4704
social_connection -0.0064 0.0726 -0.0971 0.0461 -0.0036 0.4704 1

p-values:

welfare meritocracy fatalism paternalism structural social_identity social_connection
welfare NA 0 0.0000 0.0000 0.0000 0.3527 0.7174
meritocracy 0.0000 NA 0.0000 0.0000 0.0000 0 0
fatalism 0.0000 0 NA 0.0000 0.0059 0.1882 0
paternalism 0.0000 0 0.0000 NA 0.0000 0 0.0095
structural 0.0000 0 0.0059 0.0000 NA 0.0156 0.8408
social_identity 0.3527 0 0.1882 0.0000 0.0156 NA 0
social_connection 0.7174 0 0.0000 0.0095 0.8408 0 NA

Observations

  • The social factors are mainly correlated with each other, but there does seem to be a weak positive relationship between social identity and endorsements of meritocracy and paternalism.

Individual psychological factors

We chose the framework psychological capital, which includes the variables hope, self-efficacy, resilience, and optimism. This framework represents the internal individual differences that may explain the differences in narrative endorsement.

  • hope: persevering toward goals and when necessary, redirecting paths to goals in order to succeed
  • efficacy: having confidence in one’s competence to take on and put in the necessary effort to succeed at challenging tasks
  • resilience: when beset by problems and adversity, sustaining and bouncing back and even beyond to attain success
  • optimism: a positive perception about succeeding now and in the future

And we included the framework of receptiveness to opposing views, which includes variables:

  • curiosity: the value that individuals’ place on understanding others and their views. The items in this factor reflect a desire to engage in behaviors that provide one with greater insight and information about the beliefs of others
  • derogation of opposing opinion holders: people may avoid alternative opinions based on negative judgments and attributions regarding disagreeing others and their motives
  • taboo subject: a set of beliefs about the sacred nature of some issues. Individuals high on this factor may avoid disagreement because they believe that particular issues are not subject to debate

We also included subjective wellbeing, measured using the ladder of well-being. Where we asked participants to rate their well-being ten years ago, today, and what they anticipated their wellbeing to be in five years time.

Basic stats

Note: “wellbeing” refers to where on the wellbeing ladder they would place themselves today

construct mean SE
psych_cap 3.629168 0.0118706
optimism 3.700431 0.0146773
hope 3.473020 0.0140371
efficacy 3.707794 0.0129047
resilience 3.635426 0.0133018
receptiveness 2.996754 0.0080340
curiosity 3.577532 0.0114154
derogation 3.242916 0.0125000
taboo 3.429631 0.0124317
wellbeing 5.573051 0.0376265

Correlations

r values:
Colored according to the following rules of thumb:

  • Strong Relationship: r = ±.5
  • Moderate Relationship: r = ±.3
  • Weak Relationship: r = ±.1
welfare meritocracy fatalism paternalism structural optimism hope efficacy resilience curiosity derogation taboo wellbeing
welfare 1.0000 0.5562 0.3456 0.3738 -0.3515 0.0697 0.1311 0.131 0.0989 0.0147 0.1462 0.1315 0.0905
meritocracy 0.5562 1.0000 0.1903 0.2906 -0.4387 0.147 0.2154 0.1817 0.1754 0.0458 0.1346 0.061 0.1696
fatalism 0.3456 0.1903 1.0000 0.2526 0.0489 0.0338 0.0252 0.0275 0.0261 0.0655 0.117 0.1426 -0.0333
paternalism 0.3738 0.2906 0.2526 1.0000 0.1111 0.152 0.1613 0.1349 0.1 0.1513 0.134 0.1182 0.0852
structural -0.3515 -0.4387 0.0489 0.1111 1.0000 0.0576 -0.0216 -0.0096 -0.0033 0.1755 0.0399 0.1029 -0.0645
optimism 0.0697 0.1470 0.0338 0.1520 0.0576 1 0.6752 0.6263 0.5974 0.3024 0.0333 0.0896 0.4242
hope 0.1311 0.2154 0.0252 0.1613 -0.0216 0.6752 1 0.7106 0.6529 0.2994 0.0699 0.0736 0.5091
efficacy 0.1310 0.1817 0.0275 0.1349 -0.0096 0.6263 0.7106 1 0.7243 0.3557 0.0699 0.0636 0.379
resilience 0.0989 0.1754 0.0261 0.1000 -0.0033 0.5974 0.6529 0.7243 1 0.3281 0.0696 0.0652 0.3062
curiosity 0.0147 0.0458 0.0655 0.1513 0.1755 0.3024 0.2994 0.3557 0.3281 1 -0.0363 0.0186 0.0946
derogation 0.1462 0.1346 0.1170 0.1340 0.0399 0.0333 0.0699 0.0699 0.0696 -0.0363 1 0.4603 0.0138
taboo 0.1315 0.0610 0.1426 0.1182 0.1029 0.0896 0.0736 0.0636 0.0652 0.0186 0.4603 1 0.0103
wellbeing 0.0905 0.1696 -0.0333 0.0852 -0.0645 0.4242 0.5091 0.379 0.3062 0.0946 0.0138 0.0103 1

p-values:

welfare meritocracy fatalism paternalism structural optimism hope efficacy resilience curiosity derogation taboo wellbeing
welfare NA 0.0000 0.0000 0 0.0000 0.0001 0 0 0 0.4086 0 0 0
meritocracy 0.0000 NA 0.0000 0 0.0000 0 0 0 0 0.0099 0 0.0006 0
fatalism 0.0000 0.0000 NA 0 0.0059 0.0572 0.1557 0.1212 0.1412 0.0002 0 0 0.0608
paternalism 0.0000 0.0000 0.0000 NA 0.0000 0 0 0 0 0 0 0 0
structural 0.0000 0.0000 0.0059 0 NA 0.0012 0.2242 0.5891 0.8537 0 0.0248 0 0.0003
optimism 0.0001 0.0000 0.0572 0 0.0012 NA 0 0 0 0 0.0609 0 0
hope 0.0000 0.0000 0.1557 0 0.2242 0 NA 0 0 0 0.0001 0 0
efficacy 0.0000 0.0000 0.1212 0 0.5891 0 0 NA 0 0 0.0001 0.0003 0
resilience 0.0000 0.0000 0.1412 0 0.8537 0 0 0 NA 0 0.0001 0.0002 0
curiosity 0.4086 0.0099 0.0002 0 0.0000 0 0 0 0 NA 0.0411 0.2945 0
derogation 0.0000 0.0000 0.0000 0 0.0248 0.0609 0.0001 0.0001 0.0001 0.0411 NA 0 0.4367
taboo 0.0000 0.0006 0.0000 0 0.0000 0 0 0.0003 0.0002 0.2945 0 NA 0.5628
wellbeing 0.0000 0.0000 0.0608 0 0.0003 0 0 0 0 0 0.4367 0.5628 NA

Observations

  • Several items have weak positive relationships with the harmful narratives.

Take-Aways

The strongest and notable findings are…

  • social identification has a significant positive relationship with the endorsement of meritocracy and paternalism.
  • hope has a significant positive relationship with the endorsement of welfare exploitation, meritocracy, and paternalism.
  • optimism has a significant positive relationship with the endorsement of meritocracy and paternalism.
  • efficacy has a significant positive relationship with the endorsement of welfare exploitation, meritocracy, and paternalism.
  • resilience has a significant positive relationship with the endorsement of welfare exploitation, meritocracy, and paternalism.

All of the above relationships are positive associations. This was a little confusing at first. Why would very hopeful people endorse harmful narratives like meritocracy? How could people who identify strongly with their local community also hold high paternalistic beliefs? After some collective thinking our team determined a few plausible explanations included…

  • For results like social identification, the direction of narrative endorsement would be dependent on what’s going on in your social group. If you feel like you deeply identify with your community and that community also happens to endorse harmful narratives, then you too would likely endorse harmful narratives. We see this effect play out in identity politics all the time. The high endorsement could be an effect of feeling like you identify with your local community. Perhaps those with high endorsements and similar ideologies, bond over these social values. There is a very tiny, yet significant positive association between social identification and the structural narrative. The above explanation—identification being a product of shared values—would also explain how the relationship could be positive for both harmful and helpful narratives.
  • Hopeful, optimistic people likely believe work gets you anywhere. It plays into the fetishization of working super hard as a vehicle for social mobility. It promotes order and a strong outcome expectancy that there is a 1:1 ratio between hard work and positive consequences. All this means is you firmly believe in your ability to shape your fate. Or rather, that you are confident in your capabilities and believe that your merit will get you where you want to be in life. It is the quintessential American dream and explains how one could simultaneously be hopeful, optimistic, resilient, and have high self-efficacy while also endorsing harmful narratives like meritocracy.
  • Many people are likely hopeful that if they just give others the information they need things will be better. This is what we might consider to be toxic paternalistic positivity. It makes people feel like there is a sense of control over poverty and all it takes is a little shareable information to fix all the issues. Overly simplified paths to a better future for all would be cause for hope!

Note: results broken down by meritocracy only for conciseness

Relationship w/ worldviews

Motivation: The concept and constructs that seek to quantitatively measure endorsement of narratives is relatively new. As we explore these constructs, it’s important to assess their validity. Ideologies like SDO, racial resentment, and right-wing authoritarianism are more established in the academic literature - their scales are more validated, and there is a lot of existing research on their importance as beliefs, who holds them, etc. We expect endorsements of harmful narratives to be related to beliefs in these ideologies. Showing that association helps us establish convergent validity. It furthermore gives us access to a full established literature to draw from for us (and others) to better understand our poverty narratives. These same goals are further supported by showing that these associations hold (and understanding to what extent) in this larger, newer survey we’ve conducted. The purpose of examining these worldviews is not to make the case that we should target them instead of narratives directly. Concepts like SDO, RWA, etc. are considered deeply held values. We do not have reason to expect that they would be any easier to change than poverty narratives themselves.

Approach: Below we run a series of nested models. First, examining just demographics, followed by the addition of the worldviews we’ve examined in the past (sdo, rwa, rrs), followed by the addition of worldviews we added in this survey (xenophobia & sexism). We can compare these models to assess the relative explanatory power that these worldviews have above and beyond demographics for predicting the joint (harmful) poverty narrative score. We can also look at the base demographics model to understand which demographics are more associated with harmful poverty narrative endorsement.

After removing certain responses (e.g. “Prefer not to say”, non-binary gender responses, etc.) in line with the decisions made in the Segmentation Analysis, we are left with a sample size of 2807 for this analysis.

Base model - demographics only

lm(formula = joint_pov ~ age_group + race_ethn + gender_bin + polit_ideology + education_grp + income_group + reli_imp2, data = worldviewmoddata)

  joint pov
Predictors Estimates CI p
(Intercept) 3.50 3.42 – 3.57 <0.001
age_groupAge: 45-65 -0.03 -0.08 – 0.02 0.300
age groupAge × above 65 0.01 -0.04 – 0.07 0.651
race ethn [Black +
Non-Hispanic] 		0.02 -0.05 – 0.08 0.601
race ethn [Hispanic] 0.11 0.02 – 0.20 0.022
race ethn [Other] 0.13 0.05 – 0.21 0.001
gender bin [Male] 0.13 0.08 – 0.17 <0.001
polit ideology [Moderate] -0.29 -0.35 – -0.24 <0.001
polit ideology [Somewhat
or very liberal] 		-0.63 -0.69 – -0.57 <0.001
education grp [Associate
/ Some college] 		-0.08 -0.14 – -0.02 0.009
education grp [Bachelor] -0.12 -0.18 – -0.05 0.001
education grp [Advanced] -0.20 -0.28 – -0.12 <0.001
income group [$40,000 to
$99,999] 		0.05 0.00 – 0.10 0.031
income group [$100,000 or
more] 		0.15 0.09 – 0.22 <0.001
reli imp2 [Not important] -0.13 -0.17 – -0.08 <0.001
Observations 2807
R2 / R2 adjusted 0.212 / 0.208

Worldview model

Demographics + sdo, rrs, rwa

lm(formula = joint_pov ~ age_group + race_ethn + gender_bin + polit_ideology + education_grp + income_group + reli_imp2 + rwa + rrs + sdo, data = worldviewmoddata)

  joint pov
Predictors Estimates CI p
(Intercept) 1.69 1.57 – 1.80 <0.001
age_groupAge: 45-65 -0.05 -0.09 – -0.00 0.031
age groupAge × above 65 -0.03 -0.08 – 0.01 0.165
race ethn [Black +
Non-Hispanic] 		0.18 0.13 – 0.24 <0.001
race ethn [Hispanic] 0.12 0.05 – 0.20 0.001
race ethn [Other] 0.09 0.02 – 0.15 0.007
gender bin [Male] 0.04 0.00 – 0.08 0.027
polit ideology [Moderate] -0.03 -0.07 – 0.02 0.250
polit ideology [Somewhat
or very liberal] 		-0.07 -0.13 – -0.02 0.008
education grp [Associate
/ Some college] 		-0.02 -0.07 – 0.02 0.302
education grp [Bachelor] -0.03 -0.08 – 0.03 0.330
education grp [Advanced] -0.06 -0.12 – 0.00 0.056
income group [$40,000 to
$99,999] 		0.07 0.03 – 0.11 <0.001
income group [$100,000 or
more] 		0.17 0.12 – 0.22 <0.001
reli imp2 [Not important] -0.02 -0.06 – 0.02 0.258
rwa 0.19 0.17 – 0.21 <0.001
rrs 0.23 0.21 – 0.26 <0.001
sdo 0.10 0.07 – 0.13 <0.001
Observations 2807
R2 / R2 adjusted 0.495 / 0.492

Worldview-extended model

Worldview model + sexism, xenophobia

lm(formula = joint_pov ~ age_group + race_ethn + gender_bin + polit_ideology + education_grp + income_group + reli_imp2 + rwa + rrs + sdo + sexism + xenophobia, data = worldviewmoddata)

  joint pov
Predictors Estimates CI p
(Intercept) 1.52 1.41 – 1.64 <0.001
age_groupAge: 45-65 -0.06 -0.10 – -0.01 0.008
age groupAge × above 65 -0.04 -0.09 – 0.00 0.071
race ethn [Black +
Non-Hispanic] 		0.17 0.11 – 0.22 <0.001
race ethn [Hispanic] 0.13 0.06 – 0.20 <0.001
race ethn [Other] 0.09 0.03 – 0.15 0.005
gender bin [Male] 0.04 0.00 – 0.08 0.031
polit ideology [Moderate] 0.01 -0.03 – 0.06 0.508
polit ideology [Somewhat
or very liberal] 		0.00 -0.06 – 0.06 0.987
education grp [Associate
/ Some college] 		-0.03 -0.07 – 0.02 0.283
education grp [Bachelor] -0.02 -0.07 – 0.04 0.546
education grp [Advanced] -0.04 -0.10 – 0.02 0.179
income group [$40,000 to
$99,999] 		0.07 0.03 – 0.11 0.001
income group [$100,000 or
more] 		0.17 0.12 – 0.22 <0.001
reli imp2 [Not important] -0.01 -0.05 – 0.03 0.598
rwa 0.15 0.13 – 0.17 <0.001
rrs 0.19 0.16 – 0.21 <0.001
sdo 0.08 0.05 – 0.11 <0.001
sexism 0.02 -0.00 – 0.05 0.079
xenophobia 0.11 0.08 – 0.13 <0.001
Observations 2807
R2 / R2 adjusted 0.513 / 0.510

Direct model comparisons

Analysis of Variance Table

Model 1: joint_pov ~ age_group + race_ethn + gender_bin + polit_ideology + 
    education_grp + income_group + reli_imp2
Model 2: joint_pov ~ age_group + race_ethn + gender_bin + polit_ideology + 
    education_grp + income_group + reli_imp2 + rwa + rrs + sdo
Model 3: joint_pov ~ age_group + race_ethn + gender_bin + polit_ideology + 
    education_grp + income_group + reli_imp2 + rwa + rrs + sdo + 
    sexism + xenophobia
  Res.Df    RSS Df Sum of Sq       F                Pr(>F)    
1   2792 930.04                                               
2   2789 595.55  3    334.49 540.659 < 0.00000000000000022 ***
3   2787 574.74  2     20.81  50.458 < 0.00000000000000022 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Results
Dependent variable:
joint_pov
(1) (2) (3)
age_groupAge: 45-65 -0.027 -0.046** -0.055***
(0.026) (0.021) (0.021)
age_groupAge: above 65 0.013 -0.033 -0.042*
(0.029) (0.024) (0.023)
race_ethnBlack + Non-Hispanic 0.018 0.182*** 0.168***
(0.034) (0.028) (0.027)
race_ethnHispanic 0.107** 0.124*** 0.132***
(0.047) (0.037) (0.037)
race_ethnOther 0.128*** 0.085*** 0.087***
(0.039) (0.031) (0.031)
gender_binMale 0.129*** 0.041** 0.040**
(0.023) (0.018) (0.018)
polit_ideologyModerate -0.294*** -0.026 0.015
(0.027) (0.023) (0.023)
polit_ideologySomewhat or very liberal -0.629*** -0.074*** 0.0005
(0.031) (0.028) (0.029)
education_grpAssociate / Some college -0.078*** -0.025 -0.025
(0.030) (0.024) (0.024)
education_grpBachelor -0.116*** -0.026 -0.016
(0.034) (0.027) (0.027)
education_grpAdvanced -0.201*** -0.061* -0.043
(0.040) (0.032) (0.032)
99,999 0.055** 0.074*** 0.069***
(0.025) (0.020) (0.020)
100,000 or more 0.152*** 0.170*** 0.167***
(0.033) (0.027) (0.026)
reli_imp2Not important -0.125*** -0.023 -0.010
(0.025) (0.020) (0.020)
rwa 0.188*** 0.153***
(0.010) (0.011)
rrs 0.232*** 0.189***
(0.012) (0.013)
sdo 0.100*** 0.079***
(0.015) (0.015)
sexism 0.024*
(0.013)
xenophobia 0.106***
(0.011)
Constant 3.495*** 1.687*** 1.521***
(0.037) (0.057) (0.059)
Observations 2,807 2,807 2,807
R2 0.212 0.495 0.513
Adjusted R2 0.208 0.492 0.510
Residual Std. Error 0.577 (df = 2792) 0.462 (df = 2789) 0.454 (df = 2787)
F Statistic 53.594*** (df = 14; 2792) 160.994*** (df = 17; 2789) 154.467*** (df = 19; 2787)
Note: p<0.1; p<0.05; p<0.01

Take-aways

  • The most predictive demographic characteristic for narrative beliefs by far is political ideology.
  • Unlike most demographics, age does not predict joint narrative beliefs. Neither does race, for the most part: being black doesn’t, and although there is an effect for being hispanic it is not very robust.
  • We replicate the findings that worldviews are predictive above and beyond demographics. We do not replicate that SDO is the most important one, though. Racial resentment mattered most.
  • Xenophobia was significantly predictive of narrative beliefs, even after correcting for other worldviews. Sexism was not.

Additional Requested Analyses

The analyses below are in response to ad hoc requests from the city design teams. They were run for a particular design/city, but are included below as they may nonetheless be of interest to others.

Age group differences

Some of our design work and goals are focused on voting and there is interest to understand generational differences in narrative endorsement.

Age breakdown

What is the representation of different age groups in our sample?

age_group respondents proportion
18-28 330 0.1041338
29-44 771 0.2432944
45-64 1149 0.3625749
65+ 919 0.2899968

Narrative constructs

narrative 18-28 29-44 45-64 65+
welfare 3.130909 3.170947 3.266667 3.372797
meritocracy 2.676768 2.834414 3.029011 3.119695
fatalism 3.381818 3.364462 3.245431 3.163584
paternalism 3.339394 3.273671 3.192559 3.304951
structural 3.809091 3.759533 3.632724 3.584548

Paternalism questions

DC is interested in these questions more particularly.

question 18-28 29-44 45-64 65+
Low income people could do better if someone helped them spend their money wisely. 3.172727 3.035020 2.939948 3.090316
Poor people need guidance to make better life choices. 3.318182 3.251621 3.252393 3.428727
There would be less poverty if we helped the poor plan their lives better. 3.321212 3.264591 3.142733 3.226333
We should teach poor people how to manage their finances. 3.545454 3.543450 3.435161 3.474429

Continuous data, instead of bins:

Policy intention questions

DC is interested in these questions more particularly.

question 18-28 29-44 45-64 65+
COVID assistance, like stimulus checks 3.718182 3.909209 3.532637 3.335147
Child tax credit - a tax credit given to families with dependent children 3.763636 4.059663 3.969539 3.884657
Guaranteed income program - provides a periodic unrestricted cash stipend to people’s income 3.421212 3.608301 3.220191 2.856366
Housing assistance, like temporary shelters and long-term affordable housing 3.927273 4.107652 3.987815 3.896627
Medicaid - covers healthcare costs for people with low incomes 3.990909 4.160830 4.077459 4.025027
Supplemental Nutrition Assistance Program (SNAP) - food-assistance to people with low incomes, also known as “Food Stamps” 4.042424 4.079118 4.057441 4.006529
TANF - provide temporary financial assistance for people with dependent children 3.815151 3.939040 3.817232 3.697497

Admin burden questions

DC is interested in these questions more particularly.

question 18-28 29-44 45-64 65+
Complex, detailed rules are often necessary to ensure that people are treated equally. 3.290909 3.395590 3.161010 3.133841
Food stamp programs should restrict the kind of food that you can buy. 2.621212 2.941634 3.023499 3.362350
If people want to access public services and benefits, it is only fair that they have to make a significant effort to get them. 3.075758 3.286641 3.355962 3.500544
It is acceptable that people face some hassles when they are in contact with the government. 2.878788 2.782101 2.579635 2.628944
It is acceptable that people sometimes feel that it is difficult and time-consuming to apply for government services and benefits. 3.212121 3.201038 3.104439 3.067465
People receiving unemployment benefits should prove they are actively looking for jobs. 3.624242 3.752270 3.912097 4.103373
People should be responsible for figuring out how to access government services themselves; it is not the government’s responsibility to help them. 2.578788 2.660182 2.530896 2.486398
Social assistance programs should have work requirements. 3.290909 3.457847 3.542211 3.737758
Social assistance programs should require financial counseling. 3.484849 3.623865 3.575283 3.700762
Stress and uncertainty are inevitable when people apply for government services and benefits. 3.487879 3.536965 3.496084 3.466812

GI attitude questions

DC is interested in these questions more particularly. Note: The statements on the 1-5 scale are different for these questions than most others. Be sure to reference the actual survey for their meaning.

question 18-28 29-44 45-64 65+
A guaranteed income program for my community would be… 3.766667 3.817121 3.425587 3.101197
A guaranteed income program for my life would be… 3.975758 3.964980 3.623151 3.226333
In general, do you support guaranteed income programs? 3.818182 3.726329 3.316797 2.840044

NYC Game

Ad hoc analyses to inform audience for NYC game development

We have 1331 responses from people under 50 years old (which is 42% of the full sample). Looking just among that age group:

question narrative Disagree / Strongly Disagree Neither Agree / Strongly Agree
Welfare makes people lazy. welfare 42% 25% 34%
There is a lot of fraud among welfare recipients. welfare 22% 29% 49%
An able-bodied person collecting welfare is ripping off the system. welfare 33% 29% 38%
Poor people think they deserve to be supported. welfare 33% 34% 33%
Many people take advantage of the welfare system. welfare 23% 23% 54%
Everyone has an equal opportunity to succeed. meritocracy 44% 19% 37%
We live in a meritocracy. Anyone can attain the American Dream. meritocracy 45% 23% 31%
Everyone has an equal opportunity to get a good education. meritocracy 48% 20% 32%
There will always be poor people. fatalism 9% 19% 71%
Almost by definition, someone has to be poor. fatalism 33% 37% 30%
Poverty is an inevitable outcome of society. fatalism 24% 30% 47%
Poor people need guidance to make better life choices. paternalism 24% 31% 45%
There would be less poverty if we helped the poor plan their lives better. paternalism 25% 30% 45%
Low income people could do better if someone helped them spend their money wisely. paternalism 31% 34% 35%
We should teach poor people how to manage their finances. paternalism 14% 31% 55%
Racism makes discrimination against poor minorities worse. structural 11% 20% 69%
Poor people experience prejudice and discrimination in hiring and promotion decisions at work. structural 14% 24% 61%
Poor people lack affordable housing options. structural 11% 17% 72%
Poor people lack affordable child care. structural 11% 19% 70%
Poor people lack opportunities for training & continuing education. structural 18% 22% 60%

Meritocracy endorsement is lower in this group than in older participants.

Income

Politics

Psychosocial Construct questions

When brainstorming with design teams about questions they were interested in from the national survey, a few came up related to the new psychosocial constructs we added. This section explores those questions.

Audience segments

Some of the team’s questions were about how the various audience profiles from the segmentation analysis different along the new constructs.

Do the segments identify with their communities differently?

This was probed with a Likert scale (Strongly disagree -> Strongly agree) on the following question: I identify with my local community.

segment Disagree / Strongly Disagree Neither Agree / Strongly Agree
Potential Change Agents 20% 28% 52%
Unengaged Observers 16% 44% 40%
Paternalistic Fatalists 14% 28% 58%
Suspicious Meritocrats 14% 33% 53%

Paternalistic Fatalists identify with their local community most, while Unengaged Observers identify with their local community least.

Do the segments differ in their social connectedness?

Specifically, team members wondered if Unengaged Observers are less socially connected. The motivation behind this question was to better understand this group, and to inform decisions around using community as a focus to engage disengaged people. The social connection scale used below is created by a simple average of answers (with the relevant ones reverse-coded) of the 20 social connectedness questions.